Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
Abstract This paper presents a data mining approach to meteorological and holiday values. The meteorological value
forecast the power consumption of a geographical region includes minimum temperature and humidity, maximum
based on the meteorological data. Artificial Neural temperature and humidity. The system is built with two
Network using Back Propagation is the data mining models. First one is the training model which trains the system
approach adopted here. The research is conducted using and the latter is the prediction model which forecast the
meteorological and load consumption data in Kerala required load.
region over the period of 2011- 2012. The historical data
containing temperature, humidity, public holidays and the In the training phase the preprocessed data is fed to the
daily load consumption values are used for training the system. Preprocessing the data is an important step in data
system. A predictive model is then built which gather mining as the input data greatly affects the performance of the
information from the training data set and generates system. Outlier removal and the min-max normalization are
knowledge for predicting the demand of the region. Results the two preprocessing techniques adopted. The training
on statistical significance tests assert that the proposed percentage can be selected by the user. The system will be
method can be used as an effective model to study the load trained better if more data is presented to it. The accuracy of
consumption pattern which in turn helps to forecast the the system is then checked with the remaining data known as
electricity demand of a region. the test data.
Keywords Data Mining;Artificial Neural Network;Back After the training phase the system is then ready to predict the
Propagation. load, given the required inputs. A python script is used to get
the meteorological value from the wunderground database for
I. INTRODUCTION the next 10 days. User can select the weekday for which the
demand prediction has to be made. Section 4 gives a brief
about the methodology adopted.
For best utilization of electricity, power prediction plays a
significant role in energy management. Load forecasting helps
a country to plan, schedule and manage its energy generation
to satisfy the growing needs. Forecasting demand helps to plan II. RELATED WORK
the means to satisfy the demand, efficient use of existing
energy sources, new infrastructure to be build. It also helps in There are a number of researches going on this field. Different
meeting the fluctuating demand, giving information about data mining techniques have been adopted to predict the load.
consumption pattern. Omer F. Demirel, Selim Zaim has used both SVM and neural
network to forecast electricity demand for Turkey. Hoda K.
Load forecasting is challenging because it may depend upon a Mohamed, Soliman M. El-Debeiky, Hassan M.Mahmoud,
number of factors such as the day, season, meteorological data Khaled M. El Destawy presented a model for forecasting long
and holidays. Some researches may find all data to be useful term electricity load in Egyptian Electrical Network [6]. The
but for some it may not be. Forecasting may depend on the demand is attributed to a total of eleven parameters-total
attributes that are considered. Load forecasting can be energy consumption, total energy generation, energy losses,
classified into short term, medium term, or long term. There load factor, GDP, electricity price, population, temperature,
are a number of methods for electricity forecast which humidity.
includes fuzzy logic, Artificial Neural Network (ANN),
Support Vector Machine (SVM), Regression etc. Tso[5]presents three modeling techniques for the prediction of
electricity energy consumption in two year seasons: summer
We adopt ANN with back propagation for load prediction. In and winter. The three predictive modeling techniques are
recent years, ANN as the consequence of their predominant multiple regressions, a neural network and decision tree
capability in defining the nonlinear topography between models. The base ARIMA model with just one seasonal
different nonlinear variables has become an approved pattern can be extended for the case of multiple seasons. An
procedure in the load forecasting area. We are designing a example of such an extension was presented in [3]. A
system for short term forecasting. The input considered are combinatorial problem of selecting appropriate model orders is
A neural network with back propagation momentum training Two sets of data are collected. One set consists of
algorithm was proposed in the paper Artificial Neural Network meteorological data and the other set is about the states
(ANN) trained by the Artificial Immune System (AIS) electricity consumption. The meteorological data is collected
presented by M.B. Abdul Hamid and T.K. Abdul Rahman for from the database stored by wunderground. This data consists
load forecasting in order to reduce training time and to of minimum and maximum temperature of the day, minimum
improve convergence speed for short term load forecasting and maximum humidity of the day. In addition to it the dates
model [7]. This algorithm has specific benefits such as of holidays are also collected. The electricity consumption
accuracy, speed of convergence, economic and historical data data consists of the daily demand of the state. It is collected
requirement for training etc. The major benefit of this from the Electricity Grid of the state. These data are collected
algorithm over back propagation algorithm is in terms of for the years 2011and 2012.
improvement in mean average percentage error (MAPE).
P.Sibi, S.Allwyn Jones, P.Siddarth [4] has carried out a B. Data Integration
research on different activation functions that can be applied to
the neural network.
All the data collected i.e. the electricity demand data,
meteorological data and holiday data is combined to form a
single data set. This data is combined based on the
III. ANN corresponding date.
used for back propagation converges faster with normalized with 6 nodes, two hidden layers with 4 and 2 nodes
data than with un-normalized data. Here we are using a simple respectively and an output layer with a single node (Fig 1). As
technique called Min-Max Normalization. there is no general rule to find the hidden layers it was fixed
using trial and error method. The activation function used is
log sigmoid function. It is applied to both hidden and output
layer nodes.
(4) Initialization of weights: The weights are initialized randomly
in the range 0-1. Later it gets updated to positive or negative
D. Data Mining values as the system progresses.
The rest of the input i.e. the test set is provided to the the remaining 20% of data is fed to the model for representing
prediction module and tested. Then load is predicted by the the knowledge. An x-y graph which is plotted using java is
system when providing with the necessary inputs. used for this purpose. The graph gives a plot between the
actual load and the load predicted by the model. The x axis
corresponds to the sample number and the y axis corresponds
E. Visualization of knowledge
to the load. The red colored line shows the actual load and the
green colored line shows the predicted load.
Better training occurs when more data are presented to the
system. So for training 80% of data are used. After training,
V. CONCLUSION
VI. ACKNOWLEDGMENT
REFERENCES